Automatic segmentation software in locally advanced rectal cancer: READY (REsearch program in Auto Delineation...

Area of Science:

• Radiation Oncology
• Medical Imaging Analysis

Background:

• Autocontouring software (AS) aims to improve efficiency in radiation therapy planning.
• Clinical validation is crucial to ensure the accuracy and reliability of AS for specific cancer sites like rectal cancer.

Purpose of the Study:

• To validate autocontouring software (AS) for rectal cancer delineation in a clinical setting.
• To assess the performance of AS against expert manual delineation using established quality assurance (QA) metrics.

Main Methods:

• Two radiation oncologists established reference contours for 44 rectal cancer patients.
• AS was populated with 14 cases and tested on 30 cases, comparing manual (Group A) and automatic (Group B) delineations.
• Metrics analyzed included Median Dice Similarity Coefficient (MDSC), Mean Slicewise Hausdorff Distances (MSHD), and Total-Time saving (TT).

Main Results:

• The AS achieved an MDSC of 0.75, MSHD of 2.00 mm, and 55.5% TT saving compared to manual delineation.
• Expert inter-observer agreement (MDSC) was 0.84.
• The system partially met the predefined acceptance criteria (MDSC ≥ 0.75, MSHD ≤ 1mm, TT saving ≥ 50%), with at least two criteria needing to be met, including TT saving.

Conclusions:

• Current autocontouring systems for rectal cancer offer significant time savings but do not fully meet the stringent overlap criteria for clinical validation.
• Further software development is needed to improve delineation accuracy and meet all validation benchmarks.